lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Sat, 30 Oct 2021 11:02:42 +0800
From:   Chao Yu <chao@...nel.org>
To:     Jaegeuk Kim <jaegeuk@...nel.org>
Cc:     linux-kernel@...r.kernel.org,
        linux-f2fs-devel@...ts.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH] f2fs: skip f2fs_preallocate_blocks() for
 overwrite case

On 2021/10/30 1:43, Jaegeuk Kim wrote:
> On 10/29, Chao Yu wrote:
>> Ping,
>>
>> On 2021/9/29 8:05, Chao Yu wrote:
>>> On 2021/9/29 3:08, Jaegeuk Kim wrote:
>>>> On 09/28, Chao Yu wrote:
>>>>> In f2fs_file_write_iter(), let's use f2fs_overwrite_io() to
>>>>> check whethere it is overwrite case, for such case, we can skip
>>>>> f2fs_preallocate_blocks() in order to avoid f2fs_do_map_lock(),
>>>>> which may be blocked by checkpoint() potentially.
>>>>>
>>>>> Signed-off-by: Chao Yu <chao@...nel.org>
>>>>> ---
>>>>>    fs/f2fs/file.c | 4 ++++
>>>>>    1 file changed, 4 insertions(+)
>>>>>
>>>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>>>>> index 13deae03df06..51fecb2f4db5 100644
>>>>> --- a/fs/f2fs/file.c
>>>>> +++ b/fs/f2fs/file.c
>>>>> @@ -4321,6 +4321,10 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
>>>>>            preallocated = true;
>>>>>            target_size = iocb->ki_pos + iov_iter_count(from);
>>>>> +        if (f2fs_overwrite_io(inode, iocb->ki_pos,
>>>>> +                        iov_iter_count(from)))
>>>>> +            goto write;
>>>>
>>>> This calls f2fs_map_blocks() which can be duplicate, if it's not the overwirte
>>>> case. Do we have other benefit?
>>>
>>> f2fs_overwrite_io() will break for append write case w/ below check:
>>>
>>>       if (pos + len > i_size_read(inode))
>>>           return false;
>>>
>>> I guess we may only suffer double f2fs_map_blocks() for write hole
>>> case, e.g. truncate to large size & write inside the filesize. For
>>> this case, how about adding a condition to allow double f2fs_map_blocks()
>>> only if write size is smaller than a threshold?
> 
> I still don't see the benefit much to do double f2fs_map_blocks. What is the
> problem here?

There is potential hangtask happened during swapfile's writeback:

- loop_kthread_worker_fn
  - kthread_worker_fn
   - loop_queue_work
    - lo_rw_aio
     - f2fs_file_write_iter
      - f2fs_preallocate_blocks
       - f2fs_map_blocks
        - down_read
         - rwsem_down_read_slowpath
          - schedule

I try to mitigate such issue by preallocating swapfile's block address and
avoid f2fs_do_map_lock() as much as possible in swapfile's writeback path...

Thanks,

> 
>>>
>>> Thanks,
>>>
>>>>
>>>>> +
>>>>>            err = f2fs_preallocate_blocks(iocb, from);
>>>>>            if (err) {
>>>>>    out_err:
>>>>> -- 
>>>>> 2.32.0
>>>
>>>
>>> _______________________________________________
>>> Linux-f2fs-devel mailing list
>>> Linux-f2fs-devel@...ts.sourceforge.net
>>> https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.sourceforge.net%2Flists%2Flistinfo%2Flinux-f2fs-devel&amp;data=04%7C01%7Cchao.yu%40oppo.com%7C421c06812eba4f922b0908d982dcdcc5%7Cf1905eb1c35341c5951662b4a54b5ee6%7C0%7C0%7C637684707374940190%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=u22eEWDAPaAZCyISyjTUOtQDLDuyKxTnNCI3eSwwWro%3D&amp;reserved=0

Powered by blists - more mailing lists